GROVE Framework Builds Temporally Stratified Memory from Streaming Video
Sitong Gong · hf · 2026-08-05
Researchers introduced GROVE, a training-free framework that supports both reactive QA and proactive assistance by growing memory causally from a continuous video stream.
Memory Mechanism
- GROVE retains fine-grained perceptual evidence and incrementally consolidates it into time-stamped moments, coherent episodes, and recurring cross-day patterns.
- Each stratum is paired with a scale-native retrieval skill for locating observations, replaying activities, or traversing long-range regularities.
- Reactive QA and proactive assistance share this memory and access interface, differing only in whether retrieval is initiated by a user query or the current situation.
Results
- Across multiple benchmarks including MM-lifelong and EgoServe, GROVE achieves the best results among compared methods.
- Controlled ablations show that temporal strata and their access skills are complementary, with patterns providing the largest benefit when evidence spans multiple days.
More from coding & agent
- AI Agent Autonomously Produces Mini Documentary End-to-End — illscience · 2026-08-05
- Open Source Community Slashes MiniMax H3 Video Model VRAM to 5GB in 48 Hours — ostrisai · 2026-08-05
- Building Apps in One Prompt: Testing Kimi K3 with Claude Code — markjeffrey · 2026-08-05
- AI Agent Mishandles Edit, Wiping Original X Article and Engagement — altryne · 2026-08-05
- Developer Predicts Claude Code and Codex Will Be Free Within a Year — dbasch · 2026-08-05
- NVIDIA Open-Sources CuTe Algebra and Compiler Stack to Boost AI Kernel Agents — GregoryDiamos · 2026-08-05